vLLM before 0.29.0 CPU Exhaustion via unbounded cache_salt
vllm-project vllm
Affected versions
| Product | Affected | Fixed in |
|---|---|---|
| vllm-project vllm |
|
|
Original description
vLLM versions before 0.29.0 contain a denial-of-service vulnerability in the cache_salt parameter accepted on OpenAI-compatible and Anthropic API endpoints, which lacks maximum length validation and is processed on the single EngineCore scheduler thread. Unauthenticated attackers can send HTTP requests with multi-hundred-megabyte salt values that trigger expensive pickle serialization and SHA-256 hashing, stalling the scheduler thread and denying service to all concurrent requests.
From the CVE record · VulnCheck